Online Image Information Retrieval Based On Color Affection And User Profile

As the sustained growth of web technology, the multimedia is becoming semantide that people usually used. It has become urgently which method can retrieve accurately and intelligently for multimedia information, because of the kinds of internet information.The technology of image retrieval has got great development and there are many good research fields on content-based image retrieval at present. However, it still can't meet the high accuracy. Image retrieval already becomes the focus which people pay close attention to in recent years. Then people want to search necessary information by semantic concept to get more accurate retrieval results, which can partly dispel people understand difference between low-level physical features and semantic information. This paper has done an in-depth discussion on on-line image retrieval technology, and its main work can be segmented into the following three points:First, this paper proposed a method to solve the problem that retrieving information limited, which called semantic inclusion and semantic relevant. The prerequisite of the method is to use the classification tree to sort the image information, then set up a relation from father-layer to sub-layer between each concept with the classification tree. Father-layer is a common concept of its sub-layer, and sub-layer is a detailed description to father-layer from different types, and their relation is semantic inclusion. Many sub-layers of one concept have the same father-layer, and they have the same attribute of the father, so their relation is semantic relevant.Second, this paper proposes a method to offer more accurate retrieval, which called main color clustering. This paper sets up clustering for eight types of color, such as red, green, etc. Different color can bring different affection to people. When extracting the color characteristic, each image has the record of only one main hue of those eight types of color and its affection, which can be received by separation function of color.Third, except the feedback mechanism that usually used, this paper rises a method to record and update user's individual information, which called user profile. When retrieve, system can supply more satisfying results to users by traversal the user profile. The information of the user profile can be update timely by feedback information , so as to be kept latest.This paper discussed the methods of image retrieval based on semantic information, and did experiments to prove them. The results of the experiment proved the reasonableness and effectiveness of the methods that the paper mentioned.